1. Best Practices in SQL Programming

2. Do not use irrelevant datatype  VARCHAR instead of DATETIME  CHAR(N) instead of VARCHAR(N)  etc

3. Do not use VARCHAR instead of DATETIME create table #employee_master(emp_id int, emp_name varchar(100), dob varchar(12)) insert into #employee_master select 1,'Sankar','09/19/1976' union all select 2,'Kamal','02/11/1968' union all select 3,'Rajesh','22/29/1200' select emp_name, dob from #employee_master order by dob

4. Result is emp_name dob -------------------------------------------------- Kamal 02/11/1968 Sankar 09/19/1976 Rajesh 22/29/1200

5. select emp_name, month(dob) as month_name from #employee_master Result is emp_name month_name ---------------------------------------------------------------------------------------- - Sankar 9 Kamal 2 Msg 241, Level 16, State 1, Line 1 Conversion failed when converting date and/or time from character string.

6. Forces implicit datatype convertion (from execution plan) [Expr1004] = Scalar Operator(datepart(month,CONVERT_IMPLICIT(dat etimeoffset(7),[tempdb].[dbo].[#employee_maste r].[dob],0)))

7. Wrong datatypes Force implicit convertion Store invalid data

8. Always specify size for character datatypes declare @name varchar set @name='Welcome to Chennai SQL Server User Group' select @name Result w

9. select cast('Welcome to Chennai SQL Server User Group' as varchar) Result Welcome to Chennai SQL Server

10. declare @table table (name varchar) insert into @table(name) select 'Welcome to Chennai SQL Server User Group' select name from @table

11. Msg 8152, Level 16, State 14, Line 3 String or binary data would be truncated. The statement has been terminated. (0 row(s) affected)

12. declare @v varchar set @v='124' select @v Result 1

13. What is the result? declare @v varchar set @v=124 select @v

14. declare @v varchar set @v=124 select @v Result *

15. By specifying size, you can avoid data truncation Error Unexpected result

16. CHAR(1) vs VARCHAR(1) To store Gender values like M for Male, F for Female T for True or F for False or 1 for True or 0 for False

17. Do not use VARHCAR(1) create table #test1(name char(1)) insert into #test1(name) select top 10000 'M' from sys.objects as s1 cross join sys.columns as s2 create table #test2(name varchar(1)) insert into #test2(name) select top 10000 'M' from sys.objects as s1 cross join sys.columns as s2

18. exec tempdb..sp_spaceused '#test1' exec tempdb..sp_spaceused '#test2'

19. CHAR(N) will store N bytes of data VARCHAR(N) will store actual data in bytes + extra byte to store the length of data

20. Do not use TEXT datatype  It is deprecated  String functions like left, right, etc cannot be used  No direct support of updating this column  Order by is not allowed  Union operator is not allowed

21. Do not use regional date format The regional date formats are  DD/MM/YYYY  MM/DD/YYYY  DD-MON-YYYY  etc

22. Declare @test table(dates datetime) insert into @test select '12/03/2010' select dates,month(dates) as month_value from @test Result is dates month_value ----------------------- ----------- 2010-12-03 00:00:00.000 12

23. Declare @test table(dates datetime) insert into @test select '27/03/2009' select dates,month(dates) as month_value from @test Result is Msg 242, Level 16, State 3, Line 4 The conversion of a varchar data type to a datetime data type resulted in an out-of-range value.

24. Dbcc useroptions Result is Set options Value ----------------------------------- languageus_english dateformatmdy.

25. Unambiguous formats YYYYMMDD YYYYMMDD HH:MM:SS YYYY-MM-DDTHH:MM:SS

26. set language german Declare @test table(dates datetime) insert into @test select '20100312'union all select '20090327‘ select dates from @test Result is Dates ----------------------- 2010-03-12 00:00:00.000 2009-03-27 00:00:00.000

27. set language german Declare @test table(dates datetime) insert into @test select '2010-03-12T00:00:00'union all select '2009-03-27T00:00:00‘ select dates from @test Result is Dates ----------------------- 2010-03-12 00:00:00.000 2009-03-27 00:00:00.000

28. Set dateformat dmy Declare @test table(dates datetime) insert into @test select '20100312'union all select '20090327‘ select dates from @test Result is Dates ----------------------- 2010-03-12 00:00:00.000 2009-03-27 00:00:00.000

29. Set dateformat mdy Declare @test table(dates datetime) insert into @test select '20100312'union all select '20090327‘ select dates from @test Result is Dates ----------------------- 2010-03-12 00:00:00.000 2009-03-27 00:00:00.000

30. Set dateformat ymd Declare @test table(dates datetime) insert into @test select '20100312'union all select '20090327‘ select dates from @test Result is Dates ----------------------- 2010-03-12 00:00:00.000 2009-03-27 00:00:00.000

31. Best practices when using datetime values  Always use proper DATETIME datatype to store dates and times  Do date formation at your front end application  Beware of ISDATE() function that it is not fully reliable  Always express datetime values in unambiguous formats

32. Do date formation at your front end application Declare @date datetime Set @date='20121019' Select convert(varchar(10),@date,101) Result 10/19/2012 The result is of character datatype and not useful for any date calculcations until converted back to datetime datatype

33. Don’t format dates using sql Reasons Dates become Varchars and wont allow date related caluculations (dateadd, datediff,etc) Wont allow to make use of index (if defined) if formatted at where clause Web page, reports, etc treat it as varchars (calculations, Ordering,etc wont work properly)

34. Proper usage of format function (date format in SQL)  Export results to text file with specific date format  Import data from other sources where dates are in different formats  Front end application can't be changed but it needs specific date formats for display etc

35. Efficient Querying on dates Internal Storage Two 4-byte integers First 4-bytes represents number of days from the base date (January 01, 1900) Second 4-bytes represents time which is number of milliseconds after midnight

36. Rounding Datetime value is rounded to 0.000,0.003 or 0.007 milliseconds Example select cast('2010-01-01 12:45:34.755' as datetime) Result ---------------------------------- 2010-01-01 12:45:34.757

37. Rounding numbers Last digit of the millisecond Rounded value 0 or 1 0 2,3 or 4 3 5,6,7 or 8 7 9 0 ( increase previous digit)

38. Millisecond expression select cast('2010-01-01 12:45:34.79' as datetime), cast('2010-01-01 12:45:34:79' as datetime) Result ---------------------------------------------------------------------------- 2010-01-01 12:45:34.7902010-01-01 12:45:34.080

39. Make sure millisecond part is always three digits select cast('2010-01-01 12:45:34.790' as datetime), cast('2010-01-01 12:45:34:790' as datetime) Result ----------------------------------------------------------------------------2010-01-01 12:45:34.790

40. Date '1900-01-01 00:00:00.000‘ is equal To number 0 SELECT CAST(0 as DATETIME) Result -------------------------------------- 1900-01-01 00:00:00.000 SELECT CAST(1 as DATETIME) Result -------------------------------------- 1900-01-02 00:00:00.000

41. Do not use Between operator to check datetime values declare @t table(date_col datetime) insert into @t (date_col) select '2001-01-01T12:14:20' union all select '2001-01-17T19:20:55' union all select '2001-01-21T00:00:00' union all select '2001-01-31T16:22:56' union all select '2001-02-01T00:00:00'

42. select date_col from @t where date_col between '20010101' and '20010131' Result 2001-01-01 12:14:20.000 2001-01-17 19:20:55.000 2001-01-21 00:00:00.000

43. select date_col from @t where date_col between '20010101' and '20010131 23:59:59:999' Result 2001-01-01 12:14:20.000 2001-01-17 19:20:55.000 2001-01-21 00:00:00.000 2001-01-31 16:22:56.000 2001-02-01 00:00:00.000

44. select date_col from @t where date_col >='20010101' and date_col<'20010201' Result 2001-01-01 12:14:20.000 2001-01-17 19:20:55.000 2001-01-21 00:00:00.000 2001-01-31 16:22:56.000

45. Sample Queries Find data added on yesterday Assumption : Table has a datetime column with default value getdate() Wrong method Select columns from table Where date_col=getdate()-1

46. Correct Methods Method 1 (will not make use of index) Select columns from table Where datediff(day,date_col,getdate())=1 Method 2 Select columns from table Where date_col>=dateadd(day,datediff(day,0,getdate()),-1) and date_col<dateadd(day,datediff(day,0,getdate()),0)

47. declare @orders table(order_id int identity(1,1) primary key, order_date datetime, order_value decimal(14,2)) insert into @orders(order_date,order_value) select '2001-07-16T00:52:26.913',332.38 union all select '2001-02-10T11:00:39.003',111.61 union all select '1999-01-12T13:03:42.147',265.83 union all select '1995-12-15T13:04:42.667',141.34 union all select '2000-09-10T16:09:58.817',251.30 union all select '1996-10-15T17:52:51.607',258.17 union all select '2004-11-22T10:33:05.947',222.98 union all select '2005-07-15T03:29:47.653',3148.33 union all select '2006-03-16T19:08:30.423',213.97 union all select '2006-02-21T15:38:42.770',381.91 union all select '2007-12-23T00:28:14.490',1140.51 union all select '2007-10-24T510:19:59.220',33.25

48. Find orders placed on December 15,1995 select * from @orders where order_date>='19951215' and order_date<'19951216'

49. Find orders placed from April 1,2001 to July 31, 2001 select * from @orders where order_date>='20010401' and order_date<'20010801'

50. Find orders placed on July month select * from @orders where order_id>0 and month(order_date)=7 Note : order_id>0 is used to make use of index (seek and not scan)

51. Find orders placed on or after 12 PM select * from @orders where order_id>0 and datepart(hour,order_date) >=12

52. Find orders placed in Year 2006 select * from @orders where order_date>='20060101' and order_date<'20070101‘ Do not use the following select * from @orders Where year(orderdate)=2006

53. Do not use deprecated features/functions TEXT, NTEXT datatypes – Use VARCHAR(MAX) or NVARCHAR(MAX) TIMESTAMP – Use ROWVERSION Set Rowcount option – Use TOP Operator

54. Do not use functions on indexed column create table #test(pan_no char(10) primary key) insert into #test (pan_no) select 'ALUJM8765H' union all select 'GHOKL7653K' union all select 'IMNK68765H' union all select 'LOMRF0897U' union all select 'LPIAC778J' union all select 'MZXUI1296E' Find out pan no with length =10 select pan_no from #test where len(pan_no)=10

55. Execution plan (Index scan)

56. Use Regular expression whenever needed select pan_no from #test where pan_no like '[A-Z0-9][A-Z0-9][A-Z0-9][A- Z0-9][A-Z0-9][A-Z0-9][A-Z0-9][A-Z0-9][A-Z0- 9][A-Z0-9]'

57. Execution plan (Index seek)

58. Find out Pan no with first five characters are alphabets,next four are numbers and last is an alphabet select pan_no from #test Where substring(pan_no,1,1) in (‘A’,’B’,…’Z’) substring(pan_no,2,1) in (‘A’,’B’,…’Z’) substring(pan_no,3,1) in (‘A’,’B’,…’Z’). right(pan_no,1) in (0,1,2,3,4,5,6,7,8,9)

59. select pan_no from #test where pan_no like '[A-Z][A-Z][A-Z][A-Z][A-Z][0-9][0-9][0-9][0- 9][A-Z]' Result is Pan no ---------- ALUJM8765H GHOKL7653K LOMRF0897U MZXUI1296E

60. Truncate Vs Delete Truncate is minimally logged (only page de- allocation is logged) Delete is fully logged If you want to empty the table, use always TRUNCATE

61. Beware of implicit convertion Select 5/2 (Result 2) Select 5+'2‘ (Result 7) Select 10+3.5 (Result 13.5)

62. Understand how ISNUMERIC function works Select isnumeric(12) Select isnumeric(87.4) Select isnumeric('12e2') Select isnumeric(',') Result is 1

63. Select cast(‘12e2’ as float) Result is 1200 (12*power(10,2)) Select cast(‘,’ as money) Result is 0.00 Isnumeric works for any expression that can be converted to any number datatypes

64. Missing comma and Column alias create table #employee_master(emp_id int, emp_name varchar(100), dob varchar(12)) insert into #employee_master select 1,'Sankar','09/19/1976' union all select 2,'Kamal','02/11/1968' union all select 3,'Rajesh','22/29/1200' What is the result of this query? Select emp_id, emp_name dob from #employee_master

65. emp_id dob ----------- --------- 1 Sankar 2 Kamal 3 Rajesh Check for missing comma in the SELECT statement

66. Use ORDER BY Clause only when needed create table #numbers(number int) insert into #numbers select top 100000 row_number() over (order by (select 0)) as sno from sys.columns as c1 cross join sys.columns as s2 set statistics time on select * from #numbers set statistics time off set statistics time on select * from #numbers order by number set statistics time off

67. (100000 row(s) affected) SQL Server Execution Times: CPU time = 15 ms, elapsed time = 371 ms. (100000 row(s) affected) SQL Server Execution Times: CPU time = 125 ms, elapsed time = 405 ms.

68. Use ORDER BY Clause for the following cases  Export the result in specific order to files (txt, csv, xls, etc)  Generating the serial numbers, ranking,(also resetting in each group),etc It may be useful for versions prior to 2005.  A CURSOR that should have data in specific ordering  Creating a new table (temporarily used for an analysis) with identity column from another table where identity values should depend on some ordering of some columns  Whenever TOP clause is used

69. Avoid using ORDER BY Clause for the following cases  If the result is shown in a front end application (web, reporting tool, etc), you do not need to use it in a query as it can be easily sorted in the application layer.  When adding data from one table to another. Because you still need ORDER BY clause if want results in specific order from the target table  When creating a view

70. Thank You Contact Me madhivanan2001@gmail.com Blog http://madhivanan.wordpress.com/